Application of neural network in abnormal AIS data identification

Abstract

Due to human tampering, equipment failure, channel congestion and other reasons, AIS data received by base station may have errors. These abnormal AIS data are not conducive to the identification and supervision of ship navigation intention, which greatly reduces the application value. Based on the analysis of the characteristics of the abnormal AIS data, through preprocessing and normalization of several adjacent AIS data, a model of the abnormal AIS data screening based on neural network is constructed, and the model is verified by the AIS data of the sea area near the Bohai Bay, Chengshantou Water Area, with an accuracy of 95.16%. At the same time, the influence of AIS data length and number of hidden layer nodes selected in the screening model on the accuracy rate is analyzed through experiments. The experimental results show that unreasonable data length and number of hidden layer nodes will reduce the accuracy rate of the screening model. When the data length is 4 and the number of hidden layer nodes is 6, the accuracy rate of the screening model reaches the highest.